`char-fold-table'

[Drew Adams]

Why is `char-fold-table' a defconst?

When this feature was introduced, while applauding it I
pointed to the aim of letting users and libraries define
their own equivalence classes of chars, for character
folding.

I was told to hold off, that while that is a goal, we
should not discuss it now because that might interfere
with accomplishing a preliminary version of char folding.
And that users could always modify the char table provided
or create their own, to modify the behavior.

OK.

So here we are now, with char folding.  Great.  So can
we now consider facilitating users defining their own
classes of characters?  Or making it easy for them to
modify the default equivalence classes?

Making `char-fold-table' a defconst seems wrong.  What's
the right way to enable users and code to customize such
things?

Not only do we hard-code the equivalence classes, but we
don't even tell users what they are.  Not really.  The
only thing they have so far is option - nay, defvar,
`character-fold-search', whose doc tells them only that

 "some characters will match entire groups of characters.
  For instance, " will match all variants of double quotes
  and the letter a will match all of its accented versions
  (and then some)"

"some characters"?  "For instance"?  "(and then some)"?

Yes, I know that the doc for this feature is still to be 
written, and that this the feature is still a work in
progress.  But let's please progress it - in the direction
of more and better info for users and helping users modify
and extend the behavior.